EUROPEAN COMMISSION DG RTD SEVENTH FRAMEWORK PROGRAMME THEME 7 TRANSPORT - SST SST.2011.RTD-1 GA No

Transcription

1 EUROPEAN COMMISSION DG RTD SEVENTH FRAMEWORK PROGRAMME THEME 7 TRANSPORT - SST SST.2011.RTD-1 GA No ASPECSS Assessment methodologies for forward looking Integrated Pedestrian and further extension to Cyclists Safety Deliverable No. Deliverable Title Dissemination level Written By ASPECSSD3.1 Pedestrian kinematics and specifications of new impact conditions for head- and leg-form impactors Ernesto Mottola (Toyota Motor Europe), Carmen Rodarius (TNO), SwenSchaub (TRW Automotive) April 16, 2013 Checked by Stefanie de Hair (TNO) April 19, 2013 Approved by Monica Pla (IDIADA) April 26, 2013 Issue date April 30, 2013

2 Executive summary The impactor test conditions used in regulatory and Euro NCAP testing were chosen to represent pedestrian impacts at approximately 40 km/h. AsPeCSS work package 3 (WP3) aims to define injury risk curves to assess a wider range of impact conditions, including different impact speeds, effect of vehicles representative of current European fleet and a number of other impact parameters. The goal for Task 3.1 was to provide information about the impact conditions for impactor simulations and tests by using human body model (HBM) simulation. Based on prioritized accident scenarios provided by WP1, the conditions for relevant HBM simulations were identified; because of the large amount of possible parameters to investigate and in order to leverage the advantages from the available simulation models, the task was divided in 2 subtasks; firstly a trend study was conducted with 18 simplified vehicle models from previous FP7 project Aprosys, 4 different pedestrian sizes, different pedestrian stances and orientations, and over a range of impact speed from 20 to 80kph. This first study evaluated the parameter trends and showed which were having highest influence on impact conditions. Subsequently, other partners conducted a second study with more detailed pedestrian and vehicle models in order to confirm the results from the trend study and to evaluate the effect of other parameters, such as braking and pitching. The output from these simulations was to determine the conditions for appropriate head and upper legform impactor tests, to be executed in task 3.2. Once all simulation studies were available, the results were harmonized to evaluate the headform impactor setup for different pedestrian sizes as a function of impact speed; this includes impact location, velocity and impact angle. The output of the trend study provided an overall understanding of the parameter variation, but suffered from some systematic effects induced by modeling simplifications. The detailed studies, on the other hand, provided more accurately results with a less general validity due to the limited number of vehicle models used. Results from the detailed studies were used to correct results from the trend study, and the resulting corridors were found to be in reasonable agreement to experimental data from recent literature. For upper legform impact conditions, the output of simulations with detailed models was used to investigate the best setup for an equivalent testing based on a guided EEVC impactor. The load on the pedestrian femur is increasing during the accident, and this is combined with the kinematics of the leg. The proposed approach sets the impact angle as the one where the leg suffered from maximum load and the impact velocity as the one of femur point with maximum load at the time of its contact to vehicle body. Task 3.1 studies and their output were planned and followed up with regular exchanges among the project partners; this report and the data provided fulfill the goal of the task: it enables partners from task 3.2 to investigate pedestrian injury risk based on impactor simulation and testing with realistic parameters that reproduce European accident conditions. 2/75

3 Contents Executive summary... 2 Contents Introduction Injury assessment in Work Package Objectives and plan of task T3.1: Pedestrian kinematics and specification of impact conditions Structure of the report Background information Simulation tools MADYMO HBM THUMS v Reference information for simulation setup Real world accident information APROSYS information Detailed vehicle models Approach General Trend study Model setup Simulation plan Output Detailed vehicle model study Model setup Parameter definition and variation Simulation plan Output Detailed vehicle model study Model setup Simulation plan Output Human body model simulation results Description of basic kinematics Small family car, AM50pedestrian Small family car, 6YO child pedestrian SUV, AM50 pedestrian SUV, 6YO child pedestrian Large family car, AM50 pedestrian Supermini car, AM50 pedestrian Supermini car, 6YO child pedestrian Discussion of kinematics of simplified compared to detailed model simulations /75

4 4.2 Parameter study for head impact conditions Head impact location Head impact speed Head impact angle Feedback on current Euro NCAP protocol Conditions for impactor testing setup Headform impact conditions General information Data harmonization procedure Impact conditions evaluated by trend study Head impact conditions evaluated by detailed studies Harmonized results for 6YO child and AM50 adult Comparison to PMHS test data Upper legform impact conditions Limitations of impactor testing Proposed test setup for upper leg impactor Upper leg impact conditions from detailed THUMS study Suggestions for further work Risk Register Conclusions References Acknowledgment Appendix: Tabulated results summary Headform impact conditions Upper legform impact conditions /75

5 1 Introduction 1.1 Injury assessment in Work Package 3 The objective of WP3 is to implement a method for the pedestrian injury assessment, taking into account the effectiveness of the mitigation actions from the pre-crash system, in particular vehicle speed reduction after autonomous emergency braking, or advanced passive safety measures, such as hood lifters and pedestrian airbag. WP3 activities are conducted in close interaction with WP1 where the general methodology is being defined, but considering that in WP3 this methodology defined previously will be translated into different testing, virtual (simulation)and experimental (real testing)which are meant to reflect more closely the actual accident conditions. In this Injury assessment work package, that mainly works in the in-crash phase of an accident, an objective method to assess the injury probability of a Vulnerable Road User (mainly pedestrian within the project) in case an accident with a car will be defined and evaluated. The methodology defined will consider the effectiveness of the pre-crash systems and will evaluate the enhancement on safety for the pedestrian due to the activation (or non-activation) of an integrated system. A must for this WP is to take into account existing regulatory or consumer rating procedures as basis for the methodology. Inputs to WP3 came from WP1, where accident scenarios were defined, and eventually from WP2 where active system reactions are being analysed. Such information are complemented by WP3 activities in simulation and testing defining the injury assessment method to be used developing the injury risk curves for the Cost Benefit study. Activities Task 3.1 Pedestrian kinematics and specification of impact conditions In this first task, the impact conditions of pedestrian accidents are studied by using virtual tools. Subtasks are defined as follows: Human Body Model (HBM) simulations evaluating pedestrian kinematics, based on accident scenarios and methodology from WP1 and WP2. Definition of impact conditions for setting up impactor testing based on the results of above simulations. Task 3.2 Experimental and virtual testing Evaluation of the methodology defined will be done by virtual and physical testing. Sub-tasks within this Task are: Definition of the testing methodology focusing on the in-crash phase, including tests to be done and tools to be used for the evaluation and assessment (according to general methodology from Task 1.3). Evaluation of existing regulatory or consumer testing/rating will be the basis for the methodology (EuroNCAP, regulations) Numerical simulations of impact tests (if acceptedby e.g. EuroNCAP) Preparing labs for impact conditions defined in task 3.1 Execution of test programs Validation of numerical simulations Task 3.3 Evaluation of Testing protocol The activity of this task will be focused on the final evaluation of the in crash testing and assessment. Sub-task in this Task 3.3: Evaluation of the testing protocol: New impact conditions, virtual/experimental testing elements, etc. The effects of the changed accident conditions caused by the pre-crash actions like braking, on the impacts severity of the pedestrian, will be estimated with a numerical study. The injury risk for pedestrian in a wide range of impact conditions will be estimated. The parameters under investigation during this study are impact speed, impact location, impact angle, car front geometries and car front stiffness. The results of this study will be used to specify the overall assessment methodology in WP1. This report focuses on the activities and results of task T3.1. 5/75

6 1.2 Objectives and plan of task T3.1: Pedestrian kinematics and specification of impact conditions The impactor test conditions used in regulatory and Euro NCAP testing were chosen to represent a pedestrian impact at approximately 40 km/h. It is necessary to derive different impactor test conditions to represent hitting the pedestrian at different speeds to provide the data required to derive injury risk curves. In this first task, pedestrian kinematics resulting from impacts with pedestrians at different speeds will be studied using Human Body models (HBM) to derive impactor test conditions for the upper legform and headform impactor tests. No specific information about lower leg impactor test setup were provided because the impact conditions are considered to be same as the initial impact conditions in the accident simulations. The study will be performed for scenarios identified in WP1 with relevance for forward looking integrated pedestrian safety systems. Relevant accident details like speed, impact location on the vehicle, pedestrian sizes and car models will be identified in WP1. Potential speed reductions for the given accidents will be derived from existing knowledge of on the market systems and from additional testing in WP2. This task started with a joint partners meeting, organized by TNO, in which the main specifications for the simulation studies were defined. Parameters under investigation include impact speed, impact location, impact angle, car front geometries, pedestrian size (in global terms meaning child or adult). Based on the outcome of this meeting, simulation studies were conducted by TNO, Toyota and TRW. Generic Car Models and some other specific car models were used in these Human Body Model simulations, in order to be able to assess the impact conditions for different speeds and different vehicle types, considering the big influence of the geometry of the car in the impact conditions of the pedestrian accidents. The simulation results needs to define updated impactor test conditions in terms of speed, impact angle and location for the impactor testing and simulations (head and legform) in task 3.2. Here it is important to note that for the development of the single point injury risk curves in task 3.3, it is necessary to perform tests at those points where a pedestrian head may impact. In contrast to current regulatory and consumer testing, where impact points are chosen on a worse case basis, the testing in WP3 will have to be performed on points all over the car. TNO used validated MADYMO human body models to study the pedestrian kinematics in a wide range of impact variations. For the parametric runs, the generic set-up that was recently developed to study cyclist kinematics ( simulation runs) was used. The parametric set-up includes a number of generic car front models, each with geometry variations to cover current and future vehicles geometries, and human body model of various sizes (note that the simulations were done only on the pedestrian kinematics, results of previous activities on cyclists will be forwarded to the project under task 1.4 and not repeated here). TRW used a full FE/facet MADYMO model for a specific vehicle in combination with the AM50 MADYMO HBM to investigate pedestrian kinematics. Parameters like vehicle speed, braking, pitching, as well as pedestrian posture and position were varied for this study. Furthermore, potential benefits of adding additional passive safety elements, e.g. a pedestrian airbag, were investigated. TOYOTA conducted detailed full scale simulations using the THUMS Human Body models which are based on finite element technology. The parameters for this simulation were vehicle velocity (or impacting velocity) and potential braking condition, pedestrian stance at the time of impact and pedestrian size (AM50, 6YO). Relevant vehicle models were selected based on the outcome of the WP1 accident survey. Once all simulation studies were available, the task partners harmonized the data to derive updated test specifications for the impactor tests in task 3.2. TOYOTA gathered all these data and evaluated equivalent impactor conditions. This includes impact location, velocity and impact angle as function of the impact speed, for different vehicle types. 6/75

7 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Alignment to toher workpackages and planning T3.1 Step 1 - TNO Simplified Simulations (MADYMO) WP1 Input T3.1 Step 2 - Toyota, TRW Deatiled Simulations (THUMS, MADYMO facet) Harmonization & reporting D3.1 - Pedestrian kinematics and specification of impact conditions > Accident sceanrios > Test scenarios > Prioritize parmeters T3.2 Impactor simulation T3.2 Impactor testing T3.3 Construction of injury risk curves Figure 1-1 Task 3.1 schedule 1.3 Structure of the report This report is structured in chapters as follows: Chapter 1: introduces the scope of the work package 3, and of task T3.1 in particular. The participating members and their contribution are also introduced here. Chapter 2: summarizes the background information on the simulation activities, from the available simulation technologies to the information on pedestrian accident scenarios. Chapter 3: describes the simulation approach followed by T3.1 partners. The parameters investigated and used output variables are presented here. Chapter 4: presents the results from the simulation activities and the discussion of parameter effect on pedestrian accident kinematics. Chapter 5: presents conditions to setup equivalent impactor testing; in case of headform impactor, data from different simulation tasks are harmonized to get a consistent summary; in case of upper legform impactor, data from detailed human body model simulation are used to propose a testing condition which accounts for the actual injury risk from vehicle structure. Chapter 6: suggests possible further investigations which might contribute to the scope of identifying more accurate conditions for impactor testing. Chapter 7: gathers information about identified risks and the way to manage them. Chapter 8: contains the conclusions. Chapter 9: contains the literature list. Chapter 10: acknowledges the contributions to this report. 7/75

8 2 Background information 2.1 Simulation tools The main goal of task 3.1 is to gather information on pedestrian kinematics to generate impactor test conditions for upper leg and head impactor tests to be conducted within task 3.2. To achieve this goal, a broad bandwidth of input parameters needs to be studied and evaluated with respect to their influence on the pedestrian kinematics. In general, simulations can be conducted using either finite element (FE) models or multi-body (MB) models. Both approaches have their advantages as well as their shortcomings. FE models are usually very detailed including a precise representation of all components with detailed geometry as well as actual material properties as E-modulus or yield strength. MB models on the other hand are more simplistic often using simplified characteristics of complete structures and taking into account only what really is necessary to solve the problem at hand. Upon impact all elements in an FE structures will actually deform whereas MB structures will only penetrate each other based on their contact characteristics. Therefore for FE simulations, impacts with large deformations tend to abort easily if the deformations for single elements get too high. As the state of each single element needs to be evaluated for each time step, FE set-ups additionally have very long run times. This is especially the case if a pre-impact phase needs to be simulated on top of the actual impact. Due to the high detail of the models, the output of an FE analysis also allows for significant detail and is very reliable if the model used is well validated. MB models on the other hand can have very short run times which makes them very useful for trend studies and evaluations that do not need too much detailed output. Also, an investigation of a pre-crash phase is time wise feasible. For AsPeCSS it was decided to make use of a combination of MB and FE analysis to obtain a maximum amount of information. FE analysis only would not be sufficient to investigate a large amount of input parameters and MB analysis only may not provide sufficient detail for further analysis. Therefore, first a trend study was conducted using very simplified MB models and covering a large amount of input parameter variation. Where the detail of the models used is not sufficient, more detailed MB and FE models were used for further parameter evaluation in a second step. Additionally, first data was gathered from those simulations for setting up the injury risk curves in task 3.3 and the results were used to confirm findings from the trend study. All data was then analyzed together to gather input parameters for the impactor simulations and physical tests in task MADYMO HBM MADYMO (MathematicalDYynamicMOdel), is a multi-body software package that is widely used for automotive applications for fast and accurate calculations of injury risks and safety performance. Within MADYMO, several different occupant dummy and human models as well as pedestrian models are available. Within this task, two different kinds of pedestrian models were used. As the car models used in the trend study are basic plane models, the trend study was conducted using ellipsoid pedestrian human body models (v.5.0) as provided with standard with MADYMO [10]. 4 out of the 5 models of the MADYMO ellipsoid pedestrian model family were used (see Figure 2-1): 1.) 6 year old child model 2.) 5 th percentile female model 3.) 50 th percentile male model 4.) 95 th percentile male model Figure 2-1: MADYMO ellipsoid pedestrian models. From left to right: 6 year old child, 5 th female, 50 th male, 95 th male 8/75

9 One detailed vehicle study made use of the newer and more sophisticated facet pedestrian model (see Figure 2-2). This model is currently only available as 50 th percentile male and combines the general properties of the very well validated human occupant model with a detailed representation of legs including fracture joints. All models are validated for lateral pedestrian impact. Detailed information on the development and validation of the models can be found in [6] and [18]. The main anthropometrical parameters of the pedestrians used within this study can be found in Table 2-1. Figure 2-2: MADYMO facet pedestrian model in walking position Table 2-1 Anthropometry of the MADYMO ellipsoid pedestrian models used within this study [10] Parameter 6 YO child Small female Mid-size male Large male Standing height [m] Seated height [m] Shoulder breadth [m] Knee height [m] Weight [kg] THUMS v4 The Total Human Model for Safety (THUMS) is a finite element human model [11] developed jointly by Toyota Motor Corporation and Toyota Central R&D Labs with the purpose of simulating kinematics and injury of subjects involved in car crashes. It reproduces detailed human anatomy and biomechanical properties, including skeletal system, realistic joints, stiffness of bones and skin, etc. The model was developed over the years and current version 4 includes updates [12] such as detailed organs and improvements in ligaments and in material properties. The AM50 model represents an average adult male which has a height of 175cm and a weight of 77kg (Figure 2-3a); it is available as a standing pedestrian and in a seated occupant position. Both models have a mesh size of 3 to 5mm and are composed by around 1.8 million elements. AF05 version, representing a small female, and AM95 version, representing a large male, were developed from the base AM50 model by scaling and adjusting the anthropometry. A 6 year old (6YO) child model is not yet available in version 4 therefore an older version 1 model was used within this study. Skeletal and joints system have comparable level of anatomical detail (Figure 2-3b) as version 4 and they are therefore suitable to simulate realistic kinematics; it features a relatively simplified model of the head and the internal organs and therefore it does not provide the same level of accuracy in injury risk evaluation for soft tissues. It represents a child pedestrian which has a height of 120cm and a weight of 20kg. The THUMS capability to reproduce pedestrian accidents kinematics and injuries has been validated over the years [13][14]. Applications of THUMS range from research of injury mechanisms in accidents [15] to development of vehicles for real-world safety [16].THUMS models AM50 and 6YO were used in this task to provide detailed evaluation of kinematics during pedestrian to car impact events. 9/75

10 (a) Figure 2-3 THUMS human body model with details of anatomical structure modeling. Flesh and fat are not shown; (a) AM50 adult model; (b) 6YO child model (b) 2.2 Reference information for simulation setup Real world accident information Members of WP1 analyzed several sources of accident data to define accident scenarios which are statistically more relevant in the EU (see Figure 2-4). According to their findings [1], most frequent accidents occur with a pedestrian who is crossing the road, rather than walking in same direction as the impacting vehicle. Prioritized impact conditions are summarized in Table 2-2. Figure 2-4 Accident scenarios identified by WP1 10/75

11 For what is regarding vehicle type, WP1 members combined results from an ACEA study conducted on data from 2008 with German registration data from 2009 and 2010 to identify vehicle types most often involved in accidents and prioritize the setup for simulations ( Table 2-3). Their conclusion was that one car model from the lower medium class (SFC, C-segment HB) could be recommended for all simulations. At last, WP1 also provided information about the type of pedestrian subject to injuries. According to their findings from literature studies, walking adults are most likely to get involved in accidents, followed by children running off the walkway covered by obstructions and elderly pedestrians crossing at road junctions. For each of these pedestrian types, a median value for accident speed and for typical pedestrian speed was found. These results are summarized in Table 2-4. In summary, WP1 data suggest that simulation should investigate, with different priority: Situations where the pedestrian was hit by a vehicle when crossing the road Certain vehicle classes Adult and child pedestrians, moving at different speeds A range of impact speeds, at least from 25 to 40 kph. Table 2-2 Prioritized impact scenarios Prioritized scenario Impact condition Side Obstruction Light conditions 1. S2B Crossing straight road offside no obstruction dark 2. S5A Crossing straight road nearside obstruction day 3. S3B Crossing at a junction near and offside no obstruction dark Table 2-3 Vehicle types involved in pedestrian accidents in Europe by ACEA (2008) and by Germany registration number ( ) Vehicle Categories ACEA Germany Average Supermini (SM) 40,5% 26,0% 33,3% Small family car (SFC) 23,5% 46,7% 35,1% Large family car (LFC) 11,9% 6,2% 9,0% MPV's 9,7% 7,0% 8,4% SUV's/ Vans 8,2% 8,6% 8,4% Others 6,2% 5,5% 5,8% Table 2-4 Vehicle and pedestrian impact conditions for prioritized scenarios Prioritized scenario Vehicle speed Pedestrian Pedestrian speed (Collision speed, median) 1. S2B 11.1 m/s ( 40 km/h) Elderly (65+ years) 1.2 m/s ( 4.3 km/h) 2. S5A 7.5 m/s ( 27 km/h) Child (0-17 years) 2.2 m/s ( 8 km/h) 3. S3B 7.0 m/s ( 25 km/h) Adult (18-64 years) 1.4 m/s ( 5 km/h) APROSYS information The vehicle models used for the trend study are simplified 1D models (see Figure 2-5) that consist of 8 different planes representing the most important structures of a vehicle front. They were initially developed within the European 6 th framework project APROSYS. The stiffness of the vehicle front and bonnet has been based on the average force-deflection profiles as developed within the APROSYS project [3]. The windscreen stiffness has been estimated based on windscreen impact tests performed at TNO. All stiffness s are kept the same for all investigated car fronts so the results will not be influenced by a combination of change in geometry and 11/75

12 stiffness s, but by change in geometry only. The mass of all vehicle models was set to 1300 kg based on findings from [19]. Line A Line B Line C Line D p5 BLER p4 BR p3 LBR p2 p1 p6 50º 20º 25º 40º P7 Windscreen base Ground level p8 Windscreen top Windscreen angle p9 Figure 2-5: Vehicle model geometry from APROSYS [1] (left) and example of resulting MADYMO model (right) Based on the APROSYS work, 18 vehicle contours were defined for the simulations. These 18 contours define upper and lower boundary as well as median contour of the following vehicle classes: Large Family Car (LFC) Small Family Car (SFC) Supermini (SM) Multi-Purpose Vehicle (MPV) Sports Utility Vehicle (SUV) Roadster (RS) Figure 2-6: 18 Vehicle contours used for simulations No roadster car is used for the detailed model studies. Therefore, the results from these RS profiles will be merged together with the SM profile results when being compared to the detailed vehicle model studies. The built year of the car fronts chosen from the APROSYS work varies from 1994 to 2004 with most cars from 1999 / Concerns were raised at the beginning of the project, that these car fronts might be too old to be able 12/75

13 to cover the current car fleet on the road properly. Therefore, the centerline of several new car fronts from the different vehicle classes from 2 participating OEMs were checked against the chosen profiles. It was found that those new car fronts matched the ones obtained from APROSYS still reasonably well. It can hence be assumed that the models chosen for this trend study do still cover a wide range of not only older but also recent realistic car fronts. Therefore, no additional new study to gather centerline car front profiles was conducted as previously done during APROSYS Detailed vehicle models TRW model TRW used a detailed MADYMO model, originally built by TNO and a OEM. The model represents a typical large family car. The model uses the MADYMO facet technique to represent a detailed surface, while calculation of contact forces is based on penetration with validated force deflection curves. The model was specifically developed for pedestrian impact simulations together with the MADYMO pedestrian human body models. Figure 2-7 Profile of detailed vehicle model made available by TRW compared to Aprosys profiles Figure 2-7 shows the comparison of the TRW detailed vehicle model with APROSYS large family car (LFC) shapes. The black line shows a 7-segment approximation of the detailed vehicle profile following the APROSYS approach. Compared to the LFC profiles in grey the frontend and hood region are on the upper boundary but the whole vehicle is longer and therefore the windscreen is positioned further back. This can be of influence when comparing impact regions, since the transfer from hood impact to windscreen impact obviously takes place at a significant lower WAD for the LFC than for the detailed TRW model. Furthermore, it should be mentioned here, that all runs with the simplified APROSYS models were run with no pitching when braking. On the other side, pitching was included for the runs with the detailed models. Pitching angle and lowering of the vehicle front very much depend on the brake and suspension parameters of the vehicle, but the blue line shows a typical full brake pitching shape, which is then at the lower boundary of the LFC band, see Figure /75

14 Toyota models Toyota normally relies on computer simulation for vehicle development or research projects. Some vehicle finite element models were made available for the AsPeCSS project. These models are meant to faithfully reproduce the structural behavior of the vehicle front end, subject to dynamic loading typical of pedestrian impact scenarios. In particular, those models contain a detailed geometric representation of: Bumper cover and reinforcement beam, with its energy absorbing structures Lower energy absorber Hood structure, including lock, reinforcements and hinges Cowl and louver area Engine compartment All the parts are modeled with the proper LS-Dyna material models, which are validated and refined over the years. In order to guarantee that models would fit the purpose of the AsPeCSS study, vehicle models were chosen to be representative for age (not older than 2005), size and shape. The profiles of those models, grouped in 4 vehicle classes and compared to APROSYS profiles, are shown in Figure 2-8. Small Vehicle Profiles Lower Medium (SFC) Vehicle Profiles Max Median 400 Min TME Small A 200 TME Small B TME Mini Max Median 400 Min TME Medium Upper Medium (LFC) Vehicle Profiles 2000 SUV Profiles Max 600 Max 400 Median Min 400 Median Min 200 TME Large 200 TME SUV Figure 2-8 Profiles of detailed vehicle models made available by Toyota compared to APROSYS profiles 14/75

15 It can be observed that these vehicle have a tendency to have hood profiles higher than the average of APROSYS profiles; this result is in line with the increased attention by vehicle manufacturers towards improvement of pedestrian safety over the years. 15/75

16 3 Approach 3.1 General The purpose of task T3.1 is to determine the pedestrian kinematics in case of relevant accident conditions. From the input received by WP1 and summarized in 2.2.2, some scenarios have been identified, and in particular the most likely accident condition where the pedestrian is basically crossing vehicle trajectory; however, when practically setting up a simulation of accident kinematics, more parameters have to be considered, for both the vehicle and the pedestrian behavior, as in Figure 3-1. Figure 3-1 Parameters for impact simulation setup Not all these parameters can be described in a statistically meaningful way by investigations conducted in WP1, and some of them cannot be practically addressed by available simulation technology (for example, both driver and vehicle react to forthcoming impact). Some choices were made in terms of parameter setup in the T3.1 simulation plan. First of all, the vehicle was assumed to proceed on a straight trajectory with constant speed or constant deceleration at impact, and that no driver reaction is following the impact. As for the pedestrian, only standard body types were considered: 6 years old child (6YO) small female (AF05) average adult (AM50) large adult (AM95) All pedestrians were assumed not to react to the impact. Even in case some parameters were assumed, still there was a relevant number of factors to be investigated and some prioritization and planning were needed. A full factorial study, considering all parameters would require thousands of simulations and it is therefore practically unfeasible. Considering that the purpose of this study was to determine the effect of vehicle and pedestrian parameters on impact conditions, WP3 members agreed to conduct the study in subsequent steps: A trend study, to get a qualitative understanding of the effect of a wide range of parameter variations by impact kinematics of simplified vehicle models (TNO). A more detailed study, to confirm the quantitative trends and to complete the study of remaining parameters by using more realistic vehicle and more detailed human models (TOYOTA, TRW). The approach is summarized in Table 3-1. Table 3-1 Impact parameters studied in T3.1 simulation batches Simulation batch Vehicle parameters Pedestrian parameters Speed Type Braking & pitching Type (stature & age) Posture Orientation Impact location TNO Trend X X TRW Detailed 1 X X X X 16/75

17 TOYOTA Detailed 2 X Legend: studied effect over a range; checked effect over few cases; X fixed parameter 3.2 Trend study Model setup The first step to gain insight into pedestrian kinematics was to conduct a trend study using the simplified car models as well as the MADYMO ellipsoid human body models as presented in section and 2.2.2, respectively. The following parameters were varied according to the simulation plan presented in section 3.2.2: Pedestrian size (6 year old child, 5 th female, 50 th and 95 th male) Pedestrian stance (walking: (first struck) left leg or right leg to the front, running child) Pedestrian orientation with respect to the car (-15 / 0 / 15 degrees) Vehicle shape (18 different vehicle contours based on previous APROSYS work and cyclist study) Vehicle velocity (20 to 80 km/h) As the car models chosen for this approach were merely 1-D car models with no variation in geometry in lateral direction, no variation in lateral position was performed. All combinations of pedestrian and car were chosen in a manner that the head of the pedestrian would contact the car hence the head was not allowed to miss the front structure during impact leading in a first impact with the ground. All simulations were only run up until head contact occurred. No evaluation of any ground impact was done. A general overview on the simulation set up is provided in Figure 3-2 and Figure 3-3. Based on the APROSYS findings [1]and [3] no variation was done on vehicle stiffness or friction. All simulations were carried out using the multi body (MB) code MADYMO version 7.4. From Figure 3-2 it can be seen, the arms of the pedestrian were both positioned to the front instead of swaying one arm to the front and one to the rear. This was done to increase the stability of the simulations as in this stance the arms would less likely penetrate completely through any planes of the vehicle front due to loss of contact. The influence of this change on the pedestrian kinematics was checked with an initial limited set of simulations and found negligible. Figure 3-2: from left to right: walking adult right (non-struck) leg front, walking adult left (struck) leg front, adult walking +15 degrees towards the car (top view), adult walking -15 degrees away from the car (top view) 17/75

18 Figure 3-3: Overlay of initial position of pedestrian with struck and non-struck leg to the front walking under 15 degrees towards the car for one simulation run Simulation plan Within task 3.1, TNO performed a simulation trend study with simplified car models in order to provide a general idea on the influences of various input parameters on the kinematics of the pedestrian during a car to pedestrian impact. The study was split into 3 Steps with different (full factorial) parameter variations: Step A: Step B: Step C: 4 pedestrian models (6YO child, 5 th female, 50 th and 95 th male) 18 simplified car models 5 car velocities (20 / 30/ 40 / 50 / 60 km/h) 3 additional car velocities for 6YO child and 50 th male ( 25 / 35 / 80 km/h) 1 pedestrian stance (left leg front) 1 pedestrian to car orientation (0 degrees = perpendicular to car) 1 pedestrian model (50 th percentile male) 18 simplified car models 5 car velocities (20 / 30 / 40 / 50 / 60 km/h) 2 pedestrian stances (left leg front / right leg front) 3 pedestrian to car orientations (-15 / 0 / 15 degrees) 1 pedestrian model (6YO child) 18 simplified car models 5 car velocities 3 pedestrian stances (left leg front / right leg front / running) 3 pedestrian to car orientations (-15 / 0 / 15 degrees) This matrix resulted in a total of 1710 simulations run, of which 1683 could be used for further analysis. The remaining simulations aborted due to numerical instabilities and their output was neglected. Additional to these simulations, a limited set of runs was conducted using centerlines provided by one of the OEMs upon their request. These results were used further to define input parameters for impactor simulations on a more refined and detailed car model for task 3.2. The posture used for the running child was established based on visual examples of running children found on the internet as no standardized running child posture exists so far. To investigate the influence of different running positions, an additional small set of approximately 180 simulations was run to gain more insight into the influence of different running positions. 18/75

19 3.2.3 Output As the car models used were not suitable to extract detailed information on obtainable injuries, it was decided to only extract information on the pedestrian kinematics from the trend study. The following outputs were collected per simulation: Simulation ended normally / aborted Contact time of head w.r.t. first contact (means T=0 at first contact between human and car) Linear velocities of head CoG at first contact, Location in x and z direction of head CoG at first contact (x = 0: foremost point on the car, z = 0: ground level) First car plane contacted by the head Head angle to horizon upon impact Impact speed of upper left and right leg at first contact with car Impact angle to vertical of upper left and right leg at first contact with car First car plane contacted by upper left and right leg Figure 3-4: WAD calculation based on location of head CoG The WAD cannot be obtained directly from the simulations in an automated way due to the nature of the simulation set-up. In order to gain some information on WAD, it is therefore calculated towards the location of the head CoG at first contact instead of towards the actual contact point of the head with the car(seefigure 3-4). It is known, that this way of calculating the WAD along the car contour towards the head CoG bears some error, however the obtained results are considered accurate enough for the purpose of this study. As the car models that are used are very simple, data from this investigation should be considered as a whole and not based on single data points. Though the above mentioned data is available for each simulation run, it is not recommended to use any data from single runs only, but to cluster the results in order to obtain the desired insight into trends. Even though data was gathered not only for the head but also for both, struck and non-struck side upper leg, only the head data was analyzed further in more detail. As the struck side leg is usually still in its initial position or only slightly bended towards the car upon first contact with the car, the resulting impact angle is generally 0 (measured towards the vertical) and the impact speed matches the car speed simulated. After the first contact has occurred, the leg starts rotating which cannot be taken into account for a guided impact with an upper leg impactor as used for physical testing. As redefining the output parameters for the upper leg to gain insight into the bending angle and speed of the upper leg at the point of maximum contact force (as defined in the THUMS simulations) would have required rerunning the complete study, it was decided to neglect the upper leg results from the trend study and to proceed with the upper leg results obtained from the detailed vehicle studies only for task Detailed vehicle model study Model setup The model used by TRW was originally prepared by TNO together with a OEM. It is based on the actual vehicle geometry from a finite element crash model. Using the MADYMO facet approach, all the geometry details have been taken over. The contact model used is the typical MADYMO definition based on geometrical penetration with validated force deflection curves. On the pedestrian side, the MADYMO HBM pedestrian adult facet model 19/75

20 was used. This combination fits well together and runs quite stable. There were no runs aborted. Moderate simulation run times have been observed due to the MADYMO multi body modelling approach. All the runs were setup as 90 crossing scenario from the near side (LHD), Figure 3-5 shows the model setup graphically. The t=0 initial position was defined by a distance of 220mm from the front of the vehicle to the line of CoG for the pedestrian, see Figure 3-5 on the right. This definition allows keeping the vehicle kinematics identical for different pedestrian positions. Therefore, this setup definition is quite in line with a typical hardware test setup on the track. On the other side this also leads to the fact that for corner position of the pedestrian the first impact at the leg occurs a bit after 0ms. Consequently, when the vehicle is braking, the vehicle speed at first impact to the leg for corner position of the pedestrian is already slightly reduced. Figure 3-5 Detailed model 1 setup The used MADYMO model also has an option to introduce hood lifting upon pedestrian impact. Due to the fact that there was only one contact definition available for the hood which represents the hood up configuration, no further analysis on hood lifting effects were pursued. The model does not allow any meaningful comparison of the relvant injury risk for standard position of the hood and lifted hood with respect to injury values (HIC15). On the other side, especially for head impact locations in the windscreen, the potential benefit of a pedestrian airbag was of interest. Therefore, a generic pedestrian airbag model was introduced to the vehicle model allowing comparison of pedestrian kinematics with and without pedestrian airbag. Figure 3-6 shows the situation without pedestrian airbag (left) and with pedestrian airbag (right). It should be noted here, that the dynamic effects of airbag deployment have not been studied here. The timing is set in a way, that the airbag is fully deployed and in place when the head or shoulder gets in contact. A detailed discussion on the potential of a pedestrian airbag will be done in T3.2 and reported there. 20/75

21 Figure 3-6 Impact without (left) and with pedestrian airbag (right) Parameter definition and variation The following parameters have been investigated, see also Figure 3-7: Pedestrian: o The pedestrian speed has been fixed to 5kph. This is in line with the relevant accident scenario from WP1 and the corresponsingadult test scenario from WP2. o Two different impact positions have been used, one at the corner at approximately 15% and a second one with 50% centerline impact. o Two different pedestrian postures have been used, left leg front (LLF) and right leg front (RLF). Vehicle: o The vehicle speed was varied on two levels 20kph and 40kph. o For vehicle braking the effect of braking (velocity reduction) and pitching (shape change) were investigated separately. This allows understanding the interference of these two effects. The braking level was varied between 0g (no braking) and 1g (full braking), the pitching was varied between 0 (no pitching) and 3 (typical pitching level for full braking). Figure 3-7 Parameter variation:corner impact+rlf+no brake +no pitch (top) and center impact+llf+full brake +pitch (bottom) 21/75

22 3.3.3 Simulation plan For the 5 varied parameters a full factorial simulation matrix with 32 combinations was run. This allows to do a sensitivity analysis of all output variables with respect to the input parameters. For the braking and pitching variations it is obvious, that braking without pitching and pitching without braking do not occur in real life. However, separating parameter variation in our simulation runs allowed separating the sensitivity of the output variables to both effects. So, the varying influence over speed could be observed. Table 3-2 shows the full factorial simulation matrix. Table 3-2 Detailed study 1 simulation matrix # Vspeed [km/h] Vbraking [g] Vpitch [deg] Pposition Pposture WAD [mm] IΘ [deg] Ivel [km/h] HIT [ms] centre Near front centre Near rear corner Near front corner Near rear centre Near front centre Near rear corner Near front corner Near rear centre Near front centre Near rear corner Near front corner Near rear centre Near front centre Near rear corner Near front corner Near rear centre Near front centre Near rear corner Near front corner Near rear centre Near front centre Near rear corner Near front corner Near rear centre Near front centre Near rear corner Near front corner Near rear centre Near front centre Near rear corner Near front corner Near rear Output As output variables the following entities have been analyzed: Head impact kinematics o Head impact position on the vehicle (WAD and coordinates) o Head impact angle o Head impact speed relative to the vehicle (resultant and normal to surface) o Head impact time from the start of collision Head impact injury level (HIC15) As an example the head trajectories and impact locations on the vehicle are shown in Figure /75

23 Figure 3-8: Pedestrian head trajectories (left) and impact locations on the vehicle (right) For all output variables a sensitivity analysis was done. Figure 3-9 shows a typical result for the output variable WAD for all the runs. The Pareto chart shows the significance for each individual parameter, but also for parameter combinations. The red line marks the significance for alpha=0.05. In this case, vehicle speed is the most significant parameter, followed by pitching, braking, pedestrian position and pitching. It can be seen, that the combined braking+pitching parameter is not significant in this case. However, this behavior changes over vehicle speed, as will be discussed in the result section. Figure 3-9: Pareto chart for WAD 23/75

24 3.4 Detailed vehicle model study Model setup Pedestrian The THUMS adult model was setup to be in a realistic walking position and, when possible, to match reference model used for trend study. This setup was used in previous studies [14]. Besides obvious shape differences between the THUMS and MADYMO model due to different modeling techniques, other minor differences are observed in posture (see Figure 3-10). Both models were set to a walking position which satisfies the requirements from recent EuroNCAP protocol Rev. 6.1 [5] (heel distance P = 310mm±10mm). However, the arms are close to body and both positioned frontwards for the trend study model (for the reasons see section 3.2.1) wheras for THUMS the arms were set according to a natural walking posture, so that the forward arm corresponds to backward leg and vice versa. The interaction between arms and the bonnet is sometimes found to alter kinematics. ~65deg P = 300mm The head is also positioned in a more erect stance, with the neck angle to horizontal set to 82 deg. Two versions of the adult pedestrian model were prepared, to account for both left and right leg forward. The static and dynamic friction coefficients between shoes and ground were set according to EuroNCAP protocol to 0.3. A THUMS v4 child model version is still under development, therefore an older v1 model was used for 6YO pedestrian. In case of running child, there is no widly agreed reference position to be reproduced, therefore the model was positioned to a reasonable posture (Figure 3-11). The model was not setup for bone fracture and no special output data for injury assessment were requested, being that study to be addressed by impactor simulation and testing planned in task 3.2. (a) 82deg P = 310mm Figure 3-10 Adult model setup for detailed study; (a) MADYMO model taken as reference; (b) THUMS v4 posture in T3.1 simulations. (b) Figure 3-11 Child model setup for detailed study 24/75

25 Vehicle models Vehicle models are described in section A specific setup for AsPeCSS simulations was required to represent modeling features, such as braking, and to ease post-processing of the results. Braking was set by investigating a number of vehicles of similar classes to understand (a) the pitching angle in case of full braking conditions, and (b) the pitch center. Vehicle models were made parametric to account for impact velocity and pitching as functions of linear deceleration. Pitching was therefore applied only together with braking. One of the purposes of postprocessing analyses is to detect location of impact area. This is normally described on cars by wrap around distance (WAD). With support of commercial preprocessor ANSA, vehicle models were equipped with null shells that were used to mark the WAD distances. These parts do not contribute to the vehicle behavior, but are only plot elements that follow vehicle shape, as shown in Figure Supermini Small family car SUV Figure 3-12 WAD marking on detailed vehicle models Simulation plan The parameters to be evaluated by full detail finite element simulation had to be prioritized due to the extensive computational resources needed. By taking into account the recommendations from WP1 members, a first choice was made on vehicle types and pedestrian classes to be analyzed. Then, based on screening the results of the trend study, other parameters were found to be more or less relevant as variation of impact parameters Vehicle types The criteria for selection of vehicles to be studied in detail were: 1. Statistical relevance 2. Need to investigate differentdifferent front end types Based on the above criteria, and considering that a large family car model would already be used within the other detailed vehicle study, the vehicle types to be chosen were prioritized as follows (Table 3-3): 1. Small family car (SFC) 2. SUV 3. Supermini (SM) 4. Large family car (LFC) 5. Supermini #2 (SM2) The last 2 types would be considered only in case the total number of simulations would fit the resources allocated by plan (max ~100 cases). 25/75

26 Table 3-3 Priority for studying specific vehicle classes in detailed study 2 Vehicle type Cases Priority Model available Detailed study priority Supermini (SM) 33.3% 2 TOYOTA 3, (5) Small family car (SFC) 35.1% 1 TOYOTA 1 Large family car (LFC) 9.0% 3 TOYOTA, TRW (4) SUV / Van 8.4% 4 TOYOTA 2 MPV 8.4% 5 The profiles of the detailed vehicle models, plotted according to Aprosys convention, are presented in Figure These 3 models allow to study kinematics with 3 clearly different vehicle front end shapes Profiles of Vehicle Models for Toyota detailed study Supermini Small family car SUV Figure 3-13: Profiles of the detailed vehicle models used by Toyota in detailed study Pedestrian types WP1 results provided also information on the pedestrian types most involved in accidents and their conditions. Results summarized in Table 2-4 were consolidated in two pedestrian configurations, namely: Walking AM50 adult, speed = 1.2m/s Running 6YO child, speed = 2.2m/s As for the posture, the adult model was available in walking posture as defined above. According to the results from trend study, it was found that the leg forward would change the overall body kinematics, especially resulting in alteration of head impact location (Figure 3-14). For this reason, it was decided to conduct simulations with both conditions, left and right leg forward and arms adjusted accordingly. A summary of pedestrian models to be used in the detailed study 2 is summarized in Table 3-4. Figure 3-14 preliminary results from trend study: effect of leg positioning, left front vs. right front 26/75

27 Velocity [m/s] Displacement [mm] ASPECSS D3.1 Pedestrian kinematics and specifications of impact conditions Table 3-4 Pedestrian types for detailed simulation study 2 Pedestrian Posture Moving speed THUMS Ver. AM50 Walking LLF 1.2m/s 4 Walking RLF 6YO Running LLF 2.2m/s 1 Running RLF Other parameters The velocity range was set considering the purpose of AsPeCSS to also go beyond typical Euro NCAP value of 40kph and accounting for quality of FE simulation. In case of THUMS v4, the model has been extensively validated by correlation to PMHS test executed at 40kph, but there is no proof of accuracy for much higher impact speeds. Moreover, the simulation has more issues with numerical instability at speeds higher than 60kph. Therefore it is recommended to keep the vehicle speed below 60kph. Vehicle braking, and consequent pitching were not evaluated within the trend study. Braking and pitching of the vehicle are expected to affect: the interaction of the pedestrian with the front end due to bonnet leading edge becoming lower the head impact location, due to the deceleration of the vehicle during the impact. While this effect is negligible when compared to the effect of autonomous emergency braking before the impact (large velocity reduction), still a constant deceleration of 1G might reduce final vehicle stroke by about 100mm in case of head impacting in 140ms. This large effect can occur in case of AM50, but is not expected in case of 6YO child, where the head normally impacts before 100ms in the speed range considered. Speed reduction in case of 1G constant braking Stroke reduction in case of 1G constant braking Time [ms] (a) Time [ms] Figure 3-15 Effect of constant 1G braking on vehicle kinematics: (a) speed reduction (b) stroke reduction (b) Changes of orientation of the pedestrian direction with respect to the car in a range of ±15deg were found not to significantly affect the kinematic conditions in the trend study, except for the head impact velocity in AM50 pedestrians (see section 4.2). Pedestrian orientation was therefore considered as a lower priority parameter Simulation cases Given the limit in available resources, a target of about one hundred simulations was set for this activity. It was decided to apply a full factorial approach with fewer parameters to evaluate both an average output and the influence of each parameter in specific cases. The plan agreed with other WP3 members was summarized in Table /75

28 Table 3-5 Simulation plan for detailed study 2 Item Description No. of cases Pedestrian class AM50 walking, v=1.2m/s 2 6YO child running, v=2.2m/s Pedestrian posture Left leg forward (LLF) 2 Right leg forward (RLF) Pedestrian orientation Normal to vehicle direction 1 Vehicle SMALL, Lower medium (SFC), SUV 3 Impact location Center 1 Speed 20, 30, 40, 60 4 Vehicle condition Constant speed, no pitching 2 Full braking, pitching applied Cases from full factorial study 96 Evaluation of corner effect 4 TOTAL No. of cases Output The usual evaluation approach with THUMS human body model simulation is to consider in sequence: kinematics: trajectories and velocities of body parts, relative position to vehicle, impact locations, etc. dynamics: amount of contact forces, load distribution on bones, etc. injury: strain and stress measures on bones, ligaments, organs, etc. In case of this specific study, the simulation plan required evaluation of kinematic conditions to setup equivalent impactor tests. Therefore the model was not setup with any specific injury related parameter and analysis was focusing on body kinematics (see Table 3-6). No specific output was provided for lower legform impact conditions as they are basically depending on the initial simulation setup. For the 6YO child no legform impactors are available in regulation and rating, therefore head impact conditions were the only output. Table 3-6 Summary of output contributed from T3.1 Adult Child Headform impact conditions Trend, Detailed studies Trend and detailed study 2 (THUMS) Upper legform impact conditions Detailed study 2 (THUMS) N/A Lower legform impact conditions (Same as vehicle impact conditions) N/A Head impact conditions Head impact conditions were evaluated at the time of first contact of the head with the car. The information that was extracted from each simulation are the location of the first contact point as well as the impact velocity measured at the head center of gravity and its direction, as summarized in Figure (a) Figure 3-16 Output related to head impact conditions from THUMS simulation; (a) detection of head impact by contact force, (b) head conditions at time of impact. (b) 28/75

29 The kinematic descriptors of the head center of gravity are derived directly from LS-Dyna. For this purpose a mass-less node has been set at the head CoG location and it was rigidly connected to the skull. The location of the first contact point P1 is found using the wrap around distance null shells set on the vehicle model; an example of the process for detection of first contact point is given in Figure Figure 3-17 Detection of head impact WAD in detailed study Upper leg impact conditions Upper impact conditions can be evaluated by the information available on femur kinematics and load: from the kinematics of the femur greater trochanter and the knee, the time-histories of velocity and angular position of the upper leg in vehicle reference system can be described. The orientation angle θ0 and the impact velocity angle α0 can be evaluated independently. from the bending moment at various femur locations, the maximum load condition can be evaluated and consequently the position of the leg at that time. Sample output information related to upper leg are summarized in Figure Figure 3-18 Sample output related to upper leg impact conditions from THUMS simulation 29/75

30 4 Human body model simulation results 4.1 Description of basic kinematics In this section a qualitative evaluation of pedestrian kinematics comparing multibody simulation results from the simplified trend study to the detailed facet/fe simulation results is provided. This comparison shows the degree of approximation that arises when using the ellipsoid HBM in combination with such simplified vehicle models. On the other hand, simulation output from detailed vehicle studies is strongly influenced by the shape and design characteristics of the specific car and cannot be considered of general validity. A few characteristic cases are studied in the subsections below. All simulations analyzed are set to an impact speed of 40kph with the pedestrian positioned with the left leg forward at the centerline of the vehicle. No braking or pitching is applied and the pedestrian is oriented at 90 degrees to the vehicle centerline. Relevant differences in pedestrian simulation, if any, are discussed for each case Small family car, AM50 pedestrian 0ms 10ms 20ms 30ms 40ms 50ms 60ms 70ms 30/75

31 80ms 90ms 100ms 110ms 120ms 130ms 140ms 150ms Figure 4-1 Kinematics of AM50 adult impacted by a small family car (SFC) at 40kph; comparison of simplified and detailed simulation Figure 4-1 shows the differences found between the simplified and detailed simulation set-up for an adult male impacted by a small family car (SFC) at 40kph. At the very beginning, the leg on struck side contacts the bumper and the femur starts to rotate to follow the shape of the car (time 0-30ms). This behavior is observed in similar manner for the simplified and detailed simulation. At around 40ms, the hip starts contacting the bonnet leading edge area providing a higher force to the torso, which also starts to move (time 40-60ms). Some differences start to appear at this point most likely due to the simplified representation of hood shape and stiffness in the simplified model, which leads to a different sliding behavior of the hip over the bonnet. As a consequence, the torso in the simplified simulation rotates more and causes earlier impact of the head to windshield (time ms). For the detailed model, the smooth shape of the hood allows the legs and hip to slide which causes later torso rotation and head contact. This consequently results in amore rearward head impact point location (cf. knee location at 100ms). 31/75

32 4.1.2 Small family car, 6YO child pedestrian 0ms 10ms 20ms 30ms 40ms 50ms 60ms Figure 4-2 Kinematics of a 6YO child impacted by a small family car (SFC) at 40kph; comparison of simplified and detailed simulation For the 6YO child impact, as shown in Figure 4-2, the vehicle bumper is contacting directly the hip area rather than just the leg. In general, the kinematics observed in both simulations are quite comparable. Timing and and respective impact locations of the different body parts are quite similar in the two models. However, there are some differences which can be observed: In case of the detailed vehicle model, the different shape of lower part of the bumper cover is affecting the leg kinematics (time 10-20ms), but this is not expected to affect the overall kinematics of the torso and the head. 32/75

33 The bumper cover of the detailed vehicle model is absorbing energy by local deformation but it does not cause the hip to bounce back from the car, as seen in simplified simulation (time 40-50ms). In the detailed simulation, head kinematics are determined mostly by the interaction of the torso with the hood SUV, AM50 pedestrian 0ms 10ms 20ms 30ms 40ms 50ms 60ms 70ms 33/75

34 80ms 90ms 100ms 110ms 120ms 130ms Figure 4-3 Kinematics of AM50 adult impacted by a SUV at 40kph; comparison of simplified and detailed simulation In this SUV impact, the pelvis is contacted by bonnet leading edge very soon (time 10ms) in both simulations. As already observed in the SFC simulation, Figure 4-3 shows a similar difference between simplified and detailed model: the hip contact force in the simplified model is determined by the penetration, which does not accurately represent the actual change in bonnet shape caused by deformation induced by the impact with the pedestrian. The extra force causes an increase of torso rotation (time 30-40ms). The detailed model describes one specific car with proper hood shape and deformation which causes the body to lie down and slide on the hood (time starting from 60ms). Impact with the shoulder is established before the rotation is completed and finally the head hits the louver area at a higher wrap around distance compared to the simplified simulation. 34/75

35 4.1.4 SUV, 6YO child pedestrian 0ms 10ms 20ms 30ms 40ms 50ms 60ms Figure 4-4 Kinematics of 6YO child impacted by a SUV at 40kph; comparison of simplified and detailed simulation The kinematics of 6YO child legs is quite similar, while the hip seems to bounce back from bumper for the simplified model, as observed before. The torso behavior is also influenced by the different vehicle stiffness response: the simplified one does not include a very detailed representation of the bonnet leading edge which is contacting the chest, which induces more rotation of the torso. Later head contact in the detailed model is in this specific case due to a lower bonnet height (cf. head to bonnet clearance at 30ms). 35/75

36 4.1.5 Large family car, AM50 pedestrian 0ms 10ms 20ms 30ms 40ms 50ms 60ms 70ms 36/75

37 80ms 90ms 100ms 110ms 120ms 130ms Figure 4-5 Kinematics of AM50 adult impacted by a large family car (LFC) at 40kph; comparison of simplified and detailed simulation In case of the adult model impacted by a large family car, there are some differences in kinematics of the detailed model compared to the simplified one. In the detailed simulation femur bone rupture was observed, which is triggered by the bonnet leading edge contact before 40ms. As a consequence, the load transferred to the pelvis after rupture is decreased and this causes a reduction of body torsion. For this reason the pedestrian head in the simplified model hits the hood on his back while the pedestrian in detailed model hits the hood by the shoulder. Though femur bone rupture is also possible within the ellipsoid pedestrian model, it was not observed probably due to the more round shape of the bumper area for this specific simplified car model. Similar to the case of the small family car impact, it can be observed that the detailed vehicle model allows some more sliding of the pedestrian body over the hood, resulting in higher WAD location for the head impact. 37/75

38 4.1.6 Supermini car, AM50 pedestrian 0ms 10ms 20ms 30ms 40ms 50ms 60ms 70ms 38/75

39 80ms 90ms 100ms 110ms 120ms 130ms Figure 4-6 Kinematics of AM50 adult impacted by a supermini car (SM) at 40kph; comparison of simplified and detailed simulation Figure 4-6 shows the differences found between simplified and detailed simulation in case of an adult 50 percentile male impacted by a small vehicle in the supermini segment (SM) at 40kph. Due to similar front end shape to the small family car (cf. Figure 4-1), kinematics at the beginning of impact are similar compared to that case: the leg on the struck side impacts the bumper and the femur starts to rotate to follow the shape of the car (time 0-30ms); this behavior is once again described in a similar manner by the simplified and the detailed simulation. At around 40ms, the hip starts sliding over the bonnet of detailed vehicle model which leads to an increased WAD for the head impact to the windshield Supermini car, 6YO child pedestrian 39/75

40 0ms 10ms 20ms 30ms 40ms 50ms Figure 4-7 Kinematics of 6YO child impacted by a supermini class car (SM) at 40kph; comparison of simplified and detailed simulation The remarks made for the child impacted by a small family car are still significant for the case of impact by a supermini class car Discussion of kinematics of simplified compared to detailed model simulations The simplified and detailed simulations are conducted with pedestrian models which differences in anthropometric conditions are assumed to be negligible. Both models were validated for pedestrian to car impact up to 40 km/h. The main differences between the set-ups lie in the vehicle model that were used. For the models based on APROSYS profiles, the vehicle front end is less realistic and the structural stiffness characteristics are too general to reproduce a proper panel deformation. Moreover, the spine flexibility in the ellipsoid model is modelled by two concentrated joints that can not completely capture the bending of the spine caused by leg and hip loads. These factors result in a reduced sliding of the AM50 body over the vehicle and thus in a smaller wrap around distance for the head contact compared to the detailed simulations, as described above. For the 6YO child, the reduced body mass causes generally less deformation on the vehicle front end and there are less visible differences between the simulations. 40/75

41 4.2 Parameter study for head impact conditions Head impact location For all pedestrians the head impact location rose with increasing vehicle speed. No influence of the orientation of the pedestrian towards the car could be found. The 6 year old child would hit only the bonnet of the car, never the windscreen. The 95 th percentile male would hit the bonnet in approximately 6% of all simulated conditions but only if the initial speed of the car would be 30km/h or less. 2 cases were found where this pedestrian would hit its head on the car roof, in all other cases the first impact was established on the windscreen. The 50 th percentile male would only hit a car either on the windscreen, or on the upper bonnet plane, the 5 th female results lay in-between those of the 50 th male and the 6 year old child. As expected it could be seen, that the taller the pedestrian, the higher the head would impact on a car under similar boundary conditions. When checking the influence of the impact speed on the head impact location for the 50 th percentile male, it could be seen that there is a significant increase of impacts on the windscreen when increasing the car speed from 20 to 30 km/h (simulations considered from all vehicle shapes). When increasing the car speed even further, eventually the head will not hit the bonnet at all any longer. This indicates, that the hits on the bonnet are mainly found on the upper most part of the bonnet and proceed over the windscreen base towards the middle of windscreen with increasing car speed. ended_normally [ 0 = Abnormally, 1 = Normally] car _speed orientation Average of lookup WAD SMALL 02 SFC 03 LARGE 04 MPV 05 SUV posture[1 = llf 2 = rlf 3 = run] Left leg front Right leg front CAR ID Figure 4-8: 50 th male influence of stance on average WAD per vehicle class Though the orientation of the pedestrian towards the car was not found to influence the head impact location significantly, an influence of the pedestrian stance could be established for the 50 th percentile male (see Figure 4-8). If the struck-side (left) leg was positioned to the front, the head would hit even more often on the windscreen rather than on the bonnet and an increase of the average WAD could be seen throughout all defined vehicle classes. Only for the MPV the difference was negligible. For the 6 year old child no influence is found for different walking stances, only for changing the walking to a running stance. It was found, that a running child would generally hit the bonnet lower than a walking one (see Figure 4-8). 41/75

42 Head Impact WAD [mm] ASPECSS D3.1 Pedestrian kinematics and specifications of impact conditions Sum of count 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% car speed [km/h] & posture lookup WAD posture[1 = llf 2 = rlf 3 = run] Figure 4-9: 6YO child influence of stance on WAD Data from the detailed vehicle studies is available in lesser number and hence only allows for some punctual evaluation of parameter effects for the specific cars studied.from the detailed vehicle studies no results were obtained that were contradicting to what was found in the trend study. In case of analyses conducted with THUMS in the detailed vehicle study no. 2, all results can be plotted on the same graph, without averaging, showing the influence of parameters on WAD (see Figure 4-10, with Euro NCAP testing conditions provided as reference) Head impact WAD - Detailed study Euro NCAP Adultheadform test Euro NCAP Child headform test area THUMS - 6YO - SFC - LLF THUMS - 6YO - SFC - RLF THUMS - 6YO - SMALL - LLF THUMS - 6YO - SMALL - RLF THUMS - 6YO - SUV - LLF THUMS - 6YO - SUV - RLF THUMS - AM50 - SFC - LLF THUMS - AM50 - SFC - RLF THUMS - AM50 - SMALL - LLF THUMS - AM50 - SMALL - RLF THUMS - AM50 - SUV - LLF THUMS - AM50 - SUV - RLF 0 BRAKE CONST BRAKE CONST BRAKE CONST BRAKE CONST Vehicle Impact Velocity, approximated [kph] Figure 4-10 Head impact WAD results from all THUMS simulations There is a general trend to increase WAD with impact velocity, especially in case of AM50 pedestrian. 42/75

43 Braking, which was always combined with pitching in detailed vehicle study no. 2, is mostly increasing WAD, except in case of AM50 pedestrian at very low impact speed. The effect on the 6YO is very limited due to early head contact due to bonnet leading edge impacting the torso, rather than the legs. Pedestrian posture is affecting kinematics and causes the head impact to occur at different locations. An example case is shown in Figure 4-11 (AM50, SFC, 40kph, no braking). When left leg is forward, the body has a tendency to rotate against the car and the thigh can slide over the bonnet surface more easily (T=60ms). Moreover, the arm does not directly impact the vehicle surface (T=100ms) and the resulting head impact kinematics are significantly different, with LLF pedestrian almost lying on his back when hitting the head on windshield (T=135ms). The chosen vehicle class affects the WAD mostly by amount of the body directly contacted by the bumper, as shown by the kinematics evaluation described above. If the body is struck higher, more force will be transferred to the hip, causing earlier rotation of the torso and reduced WAD; in case of low front end vehicles, the legs will rotate and the hip would possibly slide over the bonnet, resulting in slower torso rotation and larger impact WAD. In summary, it can be observed that WAD is inversely proportional to front end height, with SUV having smallest values and SFC the largest. An impact with a SUV larger than the one considered might cause a head impact at a location with even lower WAD. Time Left leg forward Right leg forward 0ms 60ms 43/75

44 100ms 135ms Figure 4-11 Effect of pedestrian initial posture on impact kinematics; example: AM50, SFC, 40kph, no brake For detailed vehicle study no. 1, pitching was investigated separately from braking. All vehicles in the field will automatically show pitching due to braking, but depending on vehicle suspension stiffness (and other vehicle parameters) the observed level of pitching can vary from car to car. Furthermore, the effect of braking is higher for lower speeds than for higher speeds, whereas pitching is simply changing the vehicle geometry for the impact independent from speed. By separating the effects of braking and pitching, it can be analyzed whether braking or pitching effect is bigger and how the two compensate each other. Starting the analysis WAD, the overall picture for different velocities was a bit fuzzy. Especially the effects of braking and pitching were not homogeneous. So, isolated analysis for different speed levels was performed. Starting at 20 km/h, Figure 4-12 shows the Pareto chart for WAD at 20 km/h. Only braking, pitching, pedestrian position and pedestrian stance have significant influence on WAD. Comparing decreases with braking and pitching influence at 20km/h it is obvious, that braking is more significant in this case. In Figure 4-13 the influence of the selected input parameters (vehicle braking, pitching, lateral pedestrian position and pedestrian stance) on the WAD of the 50th percentile pedestrian is shown for 20 km/h. It can be seen, that WAD decreases with braking, but increases with pitching. However, for the 20km/h simulations and for the given pitch angle of 3, braking has more influence than pitching. Therefore, the combined braking+pitching is decreasing WAD. This results is consistent to what has been in found independently from the other detailed study based on THUMS. Looking to the other parameters, WAD increases for corner position and for left leg rear (LLR). 44/75

45 Term ASPECSS D3.1 Pedestrian kinematics and specifications of impact conditions Pareto Chart of the Effects for 20km/h (response is WAD, Alpha = 0,05) 33,8 A C B D ACD AC AB AD ABC BC ABD BCD BD ABCD CD F actor A B C D Name V braking V pitch Pposition Pposture Effect Figure 4-12: Pareto chart for 50 th percentile male head WAD at 20 km/h Figure 4-13: influence of selected input parameters on 50th percentile male head WAD at 20 km/h The same analysis for the 40kph simulations shows some significant differences. Figure 4-14 shows the Pareto chart for 40 km/h: pitching, pedestrian stance, pedestrian position and braking have significant influence on WAD. It is obvious, that the pitching effect is significantly higher than braking here. 45/75

46 Term ASPECSS D3.1 Pedestrian kinematics and specifications of impact conditions Pareto Chart of the Effects for 40 km/h (response is WAD, Alpha = 0,05) 20,5 B C D CD A BCD BC AD ACD ABCD AB ABD BD ABC AC F actor A B C D Name V braking V pitch Pposition Pposture Effect Figure 4-14: Pareto chart for 50th percentile male head WAD at 40 km/h Looking to the individual effects in Figure 4-15 for 40 km/h it can be seen that braking is decreasing WAD whereas pitching is increasing WAD to a higher extend. So, for 40 km/h pitching is more dominant and therefore the overall effect of braking combined with pitching is increasing WAD. Looking to the other parameters in Figure 4-15 corner position and LLR are also increasing WAD. Figure 4-15: Influence of selected input parameters on 50th percentile male head WAD at 40 km/h Similar analysis was done for head impact angle and head impact velocity. The general trends as observed in the WAD analysis could also be found there. Besides the impact speed itself, braking and pitching are the most influencing parameters, with pitching being more dominant for increasing speeds Head impact speed The head impact speed is as the WAD / head impact location highly influenced by the car speed. The higher the speed of the car, the higher the head impact speed. Unlike the head impact location, also an influence of the orientation of the pedestrian can be seen when looking at the average resultant head impact speed. 46/75

47 Head Impact Velocity [kph] ASPECSS D3.1 Pedestrian kinematics and specifications of impact conditions Average of v_res [km/h] Left leg front Right leg front Left leg front Right leg front Left leg front Right leg front Left leg front Right leg front Left leg front 01 SMALL 02 SFC 03 LARGE 04 MPV 05 SUV Right leg front orientation Figure 4-16: 50 th male influence of stance and orientation on average head impact velocity per vehicle class From Figure 4-16 several conclusions can be drawn for the average head impact speed (all car velocities considered chart originating from Step B results): It is higher for small cars compared to lager cars It is higher for left (struck-side) leg front compared to right leg front It is for both stances highest if the pedestrian is heading under 15 degrees towards the car and lowest if the pedestrian is heading under 15 degrees away from the car. The difference between the vehicle classes is much less significant for the 6 year old child as this pedestrian is much smaller. Also, the average head impact speed is lower for the child compared to the average male. From the results of the detailed simulations (Figure 4-17) it was found that head impact speed for the adult is often higher than initial impact speed to vehicle. No clear effect was found from braking and pitching. 80 Head impact velocity - Detailed study 1 & Euro NCAP impact conditions FACET - AM50 - LARGE - LLF FACET - AM50 - LARGE - RLF THUMS - AM50 - SFC - LLF THUMS - AM50 - SFC - RLF THUMS - AM50 - SMALL - LLF THUMS - AM50 - SMALL - RLF THUMS - AM50 - SUV - LLF THUMS - AM50 - SUV - RLF 10 0 BRAKE CONST BRAKE CONST BRAKE CONST BRAKE CONST Vehicle Impact Velocity, approximated [kph] Figure 4-17 Head impact velocity results from detailed studies (AM50 simulations) 47/75

48 Head Impact Angel [deg] Head Impact Velocity [kph] ASPECSS D3.1 Pedestrian kinematics and specifications of impact conditions 70 Head impact velocity - Detailed study 1 & Euro NCAP impact conditions THUMS - 6YO - SFC - LLF THUMS - 6YO - SFC - RLF THUMS - 6YO - SMALL - LLF THUMS - 6YO - SMALL - RLF THUMS - 6YO - SUV - LLF THUMS - 6YO - SUV - RLF 10 0 BRAKE CONST BRAKE CONST BRAKE CONST BRAKE CONST Vehicle Impact Velocity, approximated [kph] Figure 4-18 Head impact velocity results from detailed studies (6YO simulations) Head impact angle Results from detailed vehicle model simulations show a common tendency of the head impact angle to decrease with vehicle impact speed, but no clear trend from braking/pitching or pedestrian stance. For the 6YO pedestrian, no clear trend can be highlighted, also due to impact maybe occurring on cranium or face depending on vehicle front end height and shape. 120 Head impact angle - Detailed study 1 & Euro NCAP impact conditions FACET - AM50 - LARGE - LLF FACET - AM50 - LARGE - RLF THUMS - AM50 - SFC - LLF THUMS - AM50 - SFC - RLF THUMS - AM50 - SMALL - LLF THUMS - AM50 - SMALL - RLF THUMS - AM50 - SUV - LLF THUMS - AM50 - SUV - RLF 0 BRAKE CONST BRAKE CONST BRAKE CONST BRAKE CONST Vehicle Impact Velocity, approximated [kph] 48/75

49 Head Impact Angel [deg] ASPECSS D3.1 Pedestrian kinematics and specifications of impact conditions Figure 4-19 Head impact angle results from detailed studies (AM50 simulations) 90 Head impact angle - Detailed study 1 & Euro NCAP impact conditions THUMS - 6YO - SFC - LLF THUMS - 6YO - SFC - RLF THUMS - 6YO - SMALL - LLF THUMS - 6YO - SMALL - RLF THUMS - 6YO - SUV - LLF THUMS - 6YO - SUV - RLF 10 0 BRAKE CONST BRAKE CONST BRAKE CONST BRAKE CONST Vehicle Impact Velocity, approximated [kph] Figure 4-20 Head impact angle results from detailed studies (6YO simulations) Feedback on current Euro NCAP protocol Euro NCAP assesses the pedestrian friendliness of a car by means of a certain set of impactor tests. Within this assessment, head impactor tests are performed with a child and an adult head impactor at 40 km/h considering an impact angle of 50 and 65 degrees, respectively. Impacts with the child head impactor are basically conducted within the child zone (WAD 1000 to WAD 1500) and impacts with the adult head impactor within the adult zone (WAD 1500 to WAD 2100) with the possibility to conduct impacts with the child head up to WAD 1700 under specified circumstances. From the trend study the following findings could be made with respect to those Euro NCAP initial conditions: WAD It should be noted that as mentioned before, the WAD calculation by simplified simulation models is based on the location of the head CoG position at first contact of the head with the car instead of the actual contact point itself. Therefore, the actual WAD s could be slightly higher or lower. The following minimum / maximum WAD s were gathered from all simulation (Step A, B and C) for the respective pedestrians: 6 year old child: 1017 mm 1369 mm 5 th female: 1472 mm 1819 mm 50 th male: 1553 mm 2181 mm 95 th male: 1848 mm 2479 mm The following coverage of head impacts was established from the Step A simulations for the Euro NCAP WAD s for car speeds up to 40 km/h: WAD 1250: 31% (27%) WAD 1500: 33% (33%) WAD 1700: 46% (43%) WAD 1800: 55% (54%) 49/75

50 WAD 2100: 87% (82%) Values for up to 80 km/h (hence considering all step A simulations) are provided in (), all feasible (not aborted) simulations for all different pedestrian sizes are considered. No head impact below WAD 1000 was established for any of the car shapes, no child head impact was established above and no 50 th percentile male head impact below WAD For the 6 year old child and the 50 th percentile male pedestrian the Euro NCAP WADs hence match very well. The 5 th female results form a good transition between both pedestrian sizes, though most hits are established in the adult rather than the child area. The only pedestrians that hit their head higher on a car than WAD 2100 are the 95 th percentile male in general and the 50 th percentile male for a few cases when the car speed rises above 40 km/h. It can be concluded, that pedestrians up to a size of a 50 th percentile male are well covered within the current protocol by the chosen WADs. For the 95 th percentile male, only 23% of all head impacts fall below WAD 2100 (see Figure 4-21). ended_normally [ 0 = Abnormally, 1 = Normally] car_nr pedestrian Sum of count 100% 90% % 6 70% 6 60% 50% % 30% 20% 10% 0% lookup WAD car _speed Figure 4-21: WAD distribution over car speed for 95 th percentile male If the speed of the car is 30 km/h of higher, the head of this pedestrian is likely to hit the car higher. It could hence be argued, that in order to cover also pedestrians taller than average, an increase of the maximum WAD would be beneficial Head impact speed From the Step A simulations which considered all pedestrian sizes head impact speeds were gathered. From Figure 4-22 it can be seen, that for car speeds up to and including 50km/h a head impact speed of 40 km/h covers 92% of the impacts for the 6 year old child. For all adults, the coverage is however much lower (60 to 70%). When looking into this issue in more detail it can be seen, that for the 6 year old child the head impact speed hardly ever rises above the initial car speed. Also, the first contact between head and car is always established on the bonnet for this pedestrian. This is much different for the adult occupants which are also able to hit the windscreen. 50/75

51 Sum of count 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% lookup v_res C: 6yo F: 5th M: 50th M: 95th Pedestrian Figure 4-22: head impact speed distribution per pedestrian, only car speeds up to 50 km/h considered Figure 4-23 shows the head impact speed distribution of the 50 th percentile male over the initial car speeds. Additionally, the car speeds are split by first head impact location. The following conclusions can be drawn: The higher the car speed: o the more likely the head impact speed is higher than the car speed o the more likely the head hits the windscreen rather than the bonnet head impact speeds are higher on the windscreen compared to on the bonnet head impacts on the bonnet are well covered with an impact speed of 40 km/h 99 out of 119 hit the bonnet with an impact speed not higher than 40 km/h. Considering only car speeds up to 40 km/h, a head impact speed to the bonnet of 40 km/h covers even up to 96% of the occurring impacts. To achieve similar coverage as for bonnet impacts, head impacts on the windscreen should be conducted with a higher impact speed. Only 45% (159 out of 357) head impacts occur with a speed lower or equal to 40 km/h. Rising the head impact speed on the windscreen to 50 km/h would increase the coverage to 62% (all car speeds considered). Considering only car speeds up to 40 km/h, a head impact speed to the windscreen of 40 km/h and 50 km/h covers up to 75% and 99%, respectively /75

52 middle bonnet upper bonnet windscreen upper bonnet windscreen upper bonnet windscreen upper bonnet windscreen upper bonnet windscreen ASPECSS D3.1 Pedestrian kinematics and specifications of impact conditions ended_normally [ 0 = Abnormally, 1 = Normally] Count of count 100% 90% 80% 70% 60% 50% % 30% 20% 10% 0% car _speed lookup v_res plane of first contact Figure 4-23: head impact speed of 50 th percentile male (Step B simulations) over initial car speed and first plane contacted by head In literature similar trends can be found for PMHS tests with crash conditions representing a centerline pedestrian impact at 40 km/h. [8], [9] and [11] found that the head impact speed ranged from 68% to 146%, with a tendency for lower values for bonnet impacts compared to windscreen impacts. The hypothesis that was set up in these studies is that an higher angle of the windscreen results in a higher head impact speed as the neck cannot limit the head motion to the same extend as in a bonnet impact Head impact angle From the data of the Step A simulations that included all pedestrian sizes it was found, that for SUVs, the head impact angles are in general higher than for other cars. Also, there is a significant difference in head angle for child and adult impacts as can be seen from Figure The head impact angle for all adults is in general higher than for the 6 year old child. This is also reflected within the current Euro NCAP test procedure where impacts with the child head impactor are conducted under 50 degrees and not under 65 degrees as chosen for impacts with the adult head impactor. 52/75

53 Sum of count 100% 90% 80% 70% 60% % % % % % % C: 6yo F: 5th M: 50th M: 95th lookup angle to horizon Pedestrian Figure 4-24: head impact angle (Step A simulations) per pedestrian /75

54 WAD [mm] ASPECSS D3.1 Pedestrian kinematics and specifications of impact conditions 5 Conditions for impactor testing setup 5.1 Headform impact conditions General information Assessment for head injury risk will be performed through headform impactor testing in task 3.2, with the procedures being based as a starting point on the current Euro NCAP pedestrian protocol [5]. The purpose of AsPeCSS task 3.1 was to provide a set of adapted test parameters to assess injury risk by using impactors in conditions equivalent to those evaluated by human body model simulations. One difficulty faced by partners was to define a resultant set considering the output from different simulation environments, with different parameter variations and with different degree of accuracy. A sample output related to head WAD can be seen in Figure 5-1. Trend study results were averaged for better comparison to the detailed studies: trend simulation output from each of 18 vehicle profiles were merged into 4 equivalent vehicle classes (SM, SFC, LFC, SUV/MPV). As such, one point in the chart for the trend study corresponds to the averaged output of more simulation cases Trend Study (Averaged) Detailed study 1 (All cases) 800 Detailed study 2 (All cases) Vehicle impact speed [kph] Figure 5-1: summary of the simulation output from all T3.1 studies (head impact WAD). The difficulties in summarizing those results originated from: The different (statistical) relevance of the trend study due to higher number of simulations (Trend: ~1700, Detailed study 1: 32, Detailed study 2: 96) The different parameters studied as well as their respective spread; e.g. the trend study was executed with finer sampling of vehicle impact speed (up to 80kph), while the detailed studies are only available at 20 and 40kph (Detailed study 1) and 20, 30, 40 and 60kph (Detailed study 2) The detailed studies only considered 6YO and AM50 pedestrians, compared to 6YO, AF05, AM50 and AM95 in case of the trend study. The detailed studies provide more realistic kinematics for specific car models. The trend study provides a more general overview and respective trends rather than kinematics valid for one specific car only. 54/75

55 Head impact angle [deg] Head impact speed [kph] ASPECSS D3.1 Pedestrian kinematics and specifications of impact conditions In summary, trend study results are giving more information from a qualitative viewpoint but less reliable from a quantitative perspective which is reversed for the detailed studies. The summary of simulation output for head impact speed and angle are given as reference in Figure 5-2 and Figure Trend Study (Averaged) Detailed study 1 (All cases) 0 Detailed study 2 (All cases) Vehicle impact speed [kph] Figure 5-2: summary of the simulation output from all T3.1 studies (head impact speed) Trend Study (Averaged) Detailed study 1 (All cases) 0 Detailed study 2 (All cases) Vehicle impact speed [kph] Figure 5-3: summary of the simulation output from all T3.1 studies (head impact angle). 55/75

56 5.1.2 Data harmonization procedure It was decided to introduce a procedure for result harmonization in order to summarize both the variability evaluated by the trend study and the punctual results derived from the detailed models. Task 3.1 partners agreed on the following: 1. Vehicle impact velocity is the main parameter to consider when setting impact conditions, and therefore it was chosen to represent primarily this value as independent variable. Results from impacts at 80kph were neglected as for such speed there were only a few output sets from simplified simulations and none from detailed simulations. Even if more results would be available at 80kph, the reliability of such simulation output remains to be proven (human body models are mostly validated against PMHS tests conducted at less severe impact speeds). 2. To use the results from the trend simulations to define realistic corridors for maximum and minimum value of each head impact parameter (WAD, impact speed and angle) based on the fact that a wide range of pedestrian and vehicle related factors were investigated there. The maximum and minimum corridors were evaluated by fitting the maximum and minimum values for each speed. As observed in Figure 5-1, there is a gap between WAD results in the range and the others; this is due mostly to the height difference between 6YO and AF05 pedestrians. As a consequence, it was decided to break down the results by pedestrian type. 3. To define an average tendency by linear fitting of the averages at each speed, weighted on the number of available simulation outputs at those speeds. 4. To use the results from the detailed studies to define again a tendency for each impact parameter. 5. For the pedestrian classes of 6YO running child and AM50 walking adult, available both in trend and most of the detailed studies, adjust the probability corridors by offsetting the difference between the fitting lines. This is meant to compensate the systematic differences identified above. The harmonization procedure steps 1-5 are represented in Figure 5-4. Figure 5-4: harmonization procedure for head impact conditions 56/75

57 5.1.3 Impact conditions evaluated by trend study Figure 5-5: vehicle speed effect on head impact WAD for different classes of pedestrians (trend study) Figure 5-6: vehicle speed effect on head impact velocity for different classes of pedestrians (trend study) 57/75

58 Figure 5-7: vehicle speed effect on head impact angle for different classes of pedestrians (trend study) In general, the wide variation in 6YO child head angle impact conditions can be explained by the different mechanism involved in the impact. Most vehicles have a front end which is impacting the child on the hip and abdomen (see Figure 4-2 and Figure 4-7) or even on the shoulders in case of SUV (Figure 4-4). The body is therefore accelerated much more quickly than in case of AM50 and the head is immediately rotating to impact the bonnet. The impact angle will strongly depend on the height of the bonnet leading edge as that will allow more or less rotation before the impact. A clear example of this mechanism is reproduced in Figure 5-8, for an impact at 40kph to cars with very high and very low front ends. Figure 5-8: difference in child head impact angle depending on vehicle geometry Head impact conditions evaluated by detailed studies For the detailed simulations, only data for 6YO running child and AM50 walking adult were available and evaluated for tendencies. While head impact velocity and angle were reasonably well described by a linear fitting, it was found that WAD tendency was significantly better described by a logarithmic fitting rather than by a linear one (R 2 correlation increased from 68.7% to 79.6% in case of AM50 results); that fitting suggests the tendency of head impact location to change more at a lower impact speed than at a higher one. 58/75

59 Figure 5-9: vehicle speed effect on head impact WAD for different classes of pedestrians (detailed studies) Figure 5-10: vehicle speed effect on head impact velocity for different classes of pedestrians (detailed studies) Figure 5-11: vehicle speed effect on head impact angle for different classes of pedestrians (detailed studies) Harmonized results for 6YO child and AM50 adult Main differences between simplified and detailed simulation output were found in head impact WAD, where simplified models seem to underestimate the AM50 impact location for speeds greater than 25kph, when vehicle deformation tends to be significant. This mechanism has been explained above. The opposite effect was found in case of 6YO, but it should be remarked that detailed simulations were only considering running posture. The simplified 6YO simulations results are averaged with those from child pedestrians in walking conditions which were found to result in higher WAD. 59/75

60 Figure 5-12: summary of vehicle speed effect on head impact WAD for different classes of pedestrians Trends found for impact velocity are quite consistent between simplified and detailed simulations, considering that the average results from detailed simulations almost fall in the probability corridors evaluated in the trend study (Figure 5-13). Figure 5-13: summary of vehicle speed effect on head impact velocity for different classes of pedestrians Figure 5-14: summary of vehicle speed effect on head impact angle for different classes of pedestrians Some difference can also be found for the trends of the head impact angle for the 6 year old child. The adjusted corridors are much more defined compared to the corridors from the trend study which are quite wide and basically showing not much influence of the vehicle speed at all. As previously discussed, in case of the trend study 18 different vehicle contours were considered compared to 3 for the detailed vehicle studies. Among those contours, very high front ends would generally cause low head impact angle, regardless of the impact velocity. For cars with a lower front end, the trends are similar as for an adult, though the absolute head impact angles are in general more shallow as can be seen exemplarily in Figure This effect results in less pronounced corridors for the trend study and is much less apparent in the detailed study due to the limited amount in variation of the car fronts. 60/75

61 Head impact WAD [mm] ASPECSS D3.1 Pedestrian kinematics and specifications of impact conditions Comparison to PMHS test data The harmonized head impact conditions evaluated by HBM simulation can be compared to data from available post mortem human subject (PMHS) experiments to confirm their validity. Kerrigan et al. [7][8] conducted some PMHS pedestrian impact tests against 2 sedan and a large SUV; some of the specimens had height comparable to AM50, though their mass was quite different from AM50 (Table 5-1). Table 5-1 Recent PMHS test data Reference Specimen ID Gender Height Weight Vehicle Kerrigan et al., 2007 M4 F Sedan 1 Kerrigan et al., 2007 M5 F Sedan 1 Kerrigan et al., 2007 T6 M Sedan 1 Kerrigan et al., 2009 P2 M Sedan 2 Kerrigan et al., 2009 P1 F Large SUV Kerrigan et al., 2009 P2 M Large SUV The tests were all conducted at 40kph, with established technique of supporting the pedestrian body and head by ropes that are released at the time of vehicle contact. The samples are equipped with markers to be tracked by video analysis to evaluate kinematics. Head impact WAD information is available for all the above cases and it can be compared to task 3.1 harmonized results. AM50 Head impact WAD Vehicle initial velocity [kph] Kerrigan et al., 2007 M4 F_172.9_90.6 Sedan 1 Kerrigan et al., 2007 M5 F_174.3_92.9 Sedan 1 Kerrigan et al., 2007 T6 M_179.0_87.0 Sedan 1 Kerrigan et al., 2009 P2 M_179.0_54.4 Sedan 2 Kerrigan et al., 2009 P2 M_176.0_104.2 Large SUV Kerrigan et al., 2009 P1 F_177.0_46.7 Large SUV Figure 5-15 Comparison of head impact WAD from task 3.1 harmonized results and PMHS tests Figure 5-15 shows that specimen 2007.M5, which has height very close to AM50, fits the predicted WAD corridor. The specimens 2007.M4, shorter than AM50, falls lower than predicted corridor, while the specimens 2007.T6 and 2009.P2, taller than AM50, fall higher than predicted corridor. Specimens 2009.P1 and 2009.P2 are taller than AM50 but both fall lower than predicted corridor. It should observed that these specimens were impacted against a large SUV, most probably an American model, that could be significantly larger than those analysed in task 3.1 simulations. As observed in section 4.2.1, a vehicle with a high front end, such as a large SUV, could cause a reduction of head impact WAD for AM50 pedestrians. An additional finding is that 2009.P1 and 2009.P2 have extremely different body types, yet they impact head approximately in the same location, suggesting that pedestrian mass is not a parameter of primary importance in impact kinematics. In summary, excluding the large SUV cases, the head impact location from the specimen which has closest stature to AM50 condition falls in the predicted corridor. Kerrigan suggests that specimen stature may be the most important factor to determine impact kinematics, which is in accordance to task 3.1 simulation results. 61/75

62 5.2 Upper legform impact conditions Limitations of impactor testing Injury risk for femur and hip area is commonly assessed by upper leg impactor testing developed by the working group 17 Pedestrian Safety of EEVC (European Enhanced Vehicle-safety Committee). The Euro NCAP pedestrian protocol [4] also refers to that impactor; the setup is represented in Figure Figure 5-16: upper legform impactor test setup according to EEVC regulation and Euro NCAP test protocol The setup is determined by vehicle geometry through bonnet leading edge (BLE) and bumper lead and does not consider the actual behavior of the car s front end structure when impacted by a pedestrian. Moreover, it assumes that the impactor moves along a fixed direction without rotations and as such, it neglects the rotation of the leg occurring during impact. Moreover, the regulation and Euro NCAP setup are set to target a vehicle initial velocity of 40kph. The challenge of AsPeCSS is to consider a wider range of impact conditions, including different impact speeds and the conditions likely to induce risk of leg injury. The partners in task 3.1 agreed to utilize the information from THUMS model to evaluate worst conditions for the leg: the model was setup such that it was possible to extract moment loads from femur bone, besides kinematics. T = 0 (Initial condition) T = T0 (Initial contact) T = T* (Maximum load) Figure 5-17: sample evaluation of upper leg impact conditions by THUMS pedestrian impact simulation 62/75

63 With reference to Figure 5-17, the following outputs were monitored: N*: location on femur bone subject to maximum load at T=T* T0: time of first contact of N* area to vehicle θ*: femur angle at time of maximum load T=T* Time histories such as in Figure 5-18 confirm the understanding that can be obtained when looking at the leg kinematics and load: the femur is in quite different conditions (velocity and angle ) between the time of first contact and the time of maximum load. If the test set-up would be chosen by the conditions at initial contact, the fixed impactor direction would not deform the car as the pedestrian would. On the other hand, if test set-up is chosen just by the conditions at maximum load, the impactor velocity would be significantly lower and it would result in lower car deformation and contact forces. T0 T* Velocity of N* Leg angle θ Figure 5-18 AM50 femur kinematic conditions during pedestrian impact Proposed test setup for upper leg impactor As shown above, the upper leg injury mechanism is influenced by the conditions causing maximum load on the body, that can occur in quite different conditions than the initial ones. The BLE impactor on the other hand, has fixed impact location and angle, depending on vehicle BLE height and bumper lead if chosen according to [4]. Taking into account the suggestions from Snedeker et al. [13] and Lubbe [22], the criteria set to analyze the parameters effect are: Velocity, which is set at initial impact conditions at time T0 Angle, which is set at femur maximum load Position, which should depend on vehicle geometry and location of maximum load on the femur Setting equivalent impactor mass over the speed and impact conditions considered is still an issue. Euro NCAP suggests an energy based criterion, which is based on 40kph impact speed. On the other hand, some authors suggest to use just a fixed equivalent mass of 7.5 kg for the femur [13]. The equivalent impactor mass will be briefly investigated in Task 3.2. T = 0 (Initial condition) T = T0 (Initial contact) T = T* (Maximum load) Figure 5-19: Proposed conditions for upper leg impactor setup in case of AM50 impacted by a SFC at 40kph. 63/75

64 T = 0 (Initial condition) T = T0 (Initial contact) T = T* (Maximum load) Figure 5-20: Proposed conditions for upper leg impactor setup in case of AM50 impacted by a SUV at 40kph. T = 0 (Initial condition) T = T0 (Initial contact) T = T* (Maximum load) Figure 5-21: Proposed conditions for upper leg impactor setup in case of AM50 impacted by a supermini at 40kph. Sample cases at 40kph taken from the THUMS study show how the impactor conditions would compare to actual pedestrian impact (Figure 5-19 to Figure 5-21). In case of AM50 impacted by SFC at 40kph, the location of most severe load is quite different from the BLE. An attempt to overcome the limitations of impactor positioning based on the BLE marking is proposed by Lubbe [22], who suggests to consider leg kinematics by fixing the femur rotation center around the knee and to set the impact location on a modified bonnet leading edge reference line (MLERL); Lubbe's proposal for the leg kinematics is supported by the Task 3.1 simulations findings, but the above examples show that both BLE and MLERL cannot predict well the location of maximum load due to their purely geometric definition. In summary, the output delivered by the detailed simulations recommend a setup for upper leg tests as shown in Figure 5-22, where 511mm is the distance between knee and ground in Figure 5-22 Proposed upper leg test setup 64/75

65 Upper Leg Impact Angle [deg] Upper Leg Impact Vel.[kph] ASPECSS D3.1 Pedestrian kinematics and specifications of impact conditions THUMS AM50 adult model Upper leg impact conditions from detailed THUMS study Only 48 AM50 THUMS simulations are available to setup upper leg impact conditions - 16 per vehicle type, which is not considered representative enough to define corridors and trends as done for the head impacts. The output data were averaged by speed and charted per vehicle type, with the aim of evaluating possible relations to front end shape. The results obtained from the detailed simulations with the AM50 model are summarized in Figure 5-23 to Figure Upper Leg Impact Velocity Vehicle Impact speed [kph] Figure 5-23 Upper leg impact velocity from detailed simulations SM - BLEH=757 - MLERL=757 SFC - BLEH=710 - MLERL=710 SUV - BLEH=854 - MLERL= Upper Leg Impact Angle SM - BLEH=757 - MLERL=757 SFC - BLEH=710 - MLERL=710 SUV - BLEH=854 - MLERL= Vehicle Impact speed [kph] Figure 5-24 Upper leg impact angle from detailed simulations 65/75

66 V= 60 kph V= 40 kph Upper Leg Impact Loc. [mm] ASPECSS D3.1 Pedestrian kinematics and specifications of impact conditions 350 Upper Leg Impact Location Vehicle Impact speed [kph] Figure 5-25 Upper leg impact location from detailed simulations SM - BLEH=757 - MLERL=757 SFC - BLEH=710 - MLERL=710 SUV - BLEH=854 - MLERL=776 The charts show a clear trend for impact velocity, but do not show clear trends for impact angle and location over vehicle velocity. The impact location seems to depend on BLE height, whereas the impact angle seems to be depending on the actual vehicle response to the femur loading. An interesting result is observed in the SFC at 60kph case: impact angle and location seem to be changing completely with respect to the results at 40kph and lower impact speeds. This effect can be explained by the actual vehicle behavior: the bonnet deforms enough at 60kph to cause the femur to contact a hard point in the car (Figure 5-26, condition 2) at a different location than in the 40kph impact (Figure 5-26, condition 1). T = 20 ms T = 40 ms 1 2 Figure 5-26 Different mechanism causing maximum load on the femur when impacted by a SFC at 40kph and 60kph (actual vehicle structure not shown). These results show once more how the actual mechanism causing risk of upper leg injury is mostly affected by vehicle structure rather than its shape. The trends presented here are specific for three vehicles which are recent and relatively common in EU but cannot simply be considered as representative of their class: their architecture may be significantly different from that of similar vehicles from other manufacturers. 66/75

67 6 Suggestions for further work The methodology and the results presented in this report reflect the agreement of project partners, each with their unique experiences, to proceed towards the common goal of Task 3.1: setting conditions to evaluate injury risk in a possibly more comprehensive way than suggested by current regulation and rating test protocols. The planning phase was crucial to balance each member contribution to the requested output; some activities that were discussed during the project meetings could not be implemented due to organizational, technical or feasibility constraints. On a general perspective, it would have been interesting to investigate more vehicles by detailed simulations in order to get results of more general validity. About the methodology, some more specific investigations might be suggested to get a better understanding of the risk scenarios and their possible mitigation. The output of the simulations and tests conducted within task 3.1 and 3.2 will be used to determine injury risk for pedestrians impacted by a car. Impact locations, velocities and angles as determined by HBM simulations from task 3.1 will be used for impactor tests and simulations in task 3.2. These will result in outputs like HIC that can be used to estimate injury risk. Additional information could be gathered by additional simulations with HBM looking into additional criteria rather than the standard ones. The most recent THUMS model for example also features a detailed brain model (Figure 6-1) which allows prediction of diffused axonal injuries (DAI) by evaluation of the cumulative strain damage measure (CSDM). Such damage is not so much related to linear accelerations but rather to rotational accelerations [20] which are not (and cannot be) realistically accounted for in impactor testing. Such additional evaluation is however out of the scope of this project and therefore further work will be based on already established criteria as HIC rather than possible new ones. Figure 6-1 THUMS v4 model features for injury evaluation: brain and internal organ models. 67/75

68 7 Risk Register Risk No. What is the risk Level of risk 1 WPx.x Describe here the risks!! And please refer the to the section of the text in the document level dealing with this. WP3.1 WP3.2 WP3.3 Still no fixed set on real world accident information obtained yet from WP1. Possibly resulting in a not optimal chosen set of input parameters. Simulation models are mainly validated for speeds up to 40 kph Validity of windscreen models is limited which limits the possibilities for simulations in this area (relevant for adult head impacts) Solutions to overcome the risk Give a description how to overcome the risk / give here possible solution(s) 3 Broad variation on input parameters chosen to cover wide area of possible relevant scenarios 3 Physical testing needed to check findings of simulations 2 Data for adult head will be obtained by physical testing 1 Risk level: 1 = high risk, 2 = medium risk, 3 = Low risk 68/75

69 8 Conclusions This study addressed the need to investigate impact conditions for a range of vehicle types, impact speed and pedestrian types, while considering the scatter caused by other parameters, such as the pedestrian posture or the vehicle braking. A large number of parameters was studied by combining a trend study with simplified models and detailed studies with detailed vehicle models to confirm the trends. Results harmonization was also established by means of a comparison of the pedestrian kinematics to confirm a systematic effect from assumptions in the simplified models on the head impact conditions. From the harmonized results, equivalent test setup for headform and upper leg impactor were evaluated. The study highlighted the following trends with regards to head impact conditions: WAD is mainly depending on vehicle speed and pedestrian height. It is also influenced by the car front end shape mostly due to hip engagement mechanism, so that lower front end vehicles allow the leg to rotate and the hip to slide over the vehicle body. Larger vehicles, such as SUVs, can directly impact hip with bonnet edge and cause a faster rotation of the torso, resulting in shorter WAD. Braking with pitching were found to cause an increase in WAD, similarly to a reduction of vehicle front end, at least for impact speed greater than 20kph. The pedestrian stance was found to affect WAD mostly by the position of first impacted leg, while small differences in pedestrian orientation were not found to cause substantial changes in impact location. WAD for 6YO child pedestrian is less sensitive to vehicle type, as the impact occurs mostly at the hip or on the abdomen. Head impact velocity is also primarily depending on vehicle speed. In this case, however, some dependency was found on pedestrian stance. From the results of the trend study, it can be observed some tendency of head impact velocity to be larger for smaller vehicles. From the results of the detailed simulations it was found that head impact speed for the adult pedestrian is often higher than the initial impact speed to vehicle. No clear effect was found from braking and pitching. Head impact angle is generally depending on vehicle impact speed, with higher speeds causing lower impact angles. For the AM50 pedestrian, the head impact angle appears to depend on front end height, with SUV causing more torso rotation and therefore significantly higher impact angles. For cars which allow more sliding over the hood, windshield inclination seems also to be affecting the impact conditions: for the cases where impact is occurring on the windshield, cars with more vertical windshield (such as the supermini) would be impacted before head could rotate around the neck, resulting in lower impact angles. For the 6YO child, a large spread was found, mainly due to the different parts of the head impacting the car: for cars with high front end, the face rather than the skull may be impacting, resulting in low impact angle. With regards to 50th percentile male and 6 year old child, the results for the head impact conditions also confirm the general validity of the current Euro NCAP setup, which is based solely on a vehicle impact speed of 40kph. For upper legform impact conditions, the output of simulations with detailed AM50 models was used to investigate the best setup for an equivalent testing, based on a guided EEVC impactor. The load on the pedestrian femur is increasing during the impact event, and this is combined with the kinematics of the leg. The proposed approach sets equivalent impactor conditions as follows: Impact location was set by point N*, that is the femur bone location subject to maximum load. The impact location seems to depend on BLE height, however maximum load is actually caused by the underlying structure of the vehicle, which is not necessarily depending on its shape. Impact velocity was set as the relative velocity of point N* at the time of its first contact to the vehicle. This is clearly found to increase with vehicle impact speed. Impact direction was set as perpendicular to the angle in which femur is subject to maximum load. Also the femur angle at maximum load will depend on vehicle structure rather than on its geometry. High front end vehicles such as SUVs will be more likely to cause high load towards the hip at early stage, resulting in reduced impact angle. These quantities were evaluated by three vehicles only and therefore they should not be considered to have general validity. The impactor equivalent mass was not defined in this task and it was therefore left to the following. 69/75

70 Task 3.1 studies and their output were planned and followed up with regular exchanges among the project partners; this report and the data provided fulfill the goal of the task: they enable partners from task 3.2 to investigate pedestrian injury risk based on impactor simulation and testing with realistic parameters that reproduce European accident conditions. 70/75

71 9 References [1] ASPECSSD1.1, Scenarios and weighting factors for pre crash assessment of integrated pedestrian safety systems, AsPeCSS Project public deliverable [2] AP-SP33-022R, Concepts for a hybrid test procedure, Integrated Project APROSYS 6 th framework program, 2006 [3] AP-SP31-009R, Stiffness corridors for the current European Fleet,, Integrated Project APROSYS 6 th framework program, 2008 [4] Euro NCAP, Pedestrian Testing Protocol v5.3.1, November 2011 [5] Euro NCAP, Pedestrian Testing Protocol, v6.2.1, February 2013 [6] Hoof, J. van, Lange, R. de, Wismans, J., Improving Pedestrian safety using Numerical Human Models, STAPP Conference Proceedings, 2003 [7] Kerrigan, J., Crandall, J., Deng, b., Pedestrian kinematic response to mid-sized vehicle impact, Int. J. Of Vehicle Safety, 2007 [8] Kerrigan,J., Arregui, C., Crandall, J., Pedestrian head impact dynamics: comparison of dummy and PMHS in small sedan and large SUV impacts, ESV 2009 [9] Kerrigan, J., Crandall, J., Deng,B., A comparative analysis of pedestrian injury risk predicted by mechanical impatcors and post mortem human surrogates,stappcar Crash J.2008 [10] MADYMO Human Models Manual, Release 7.4, November 2011 [11] THUMS AM50 Pedestrian/Occupant model manual, Version 4.0, October 2011 [12] Fressmann, D., Vehicle Safety using the THUMS Human model, German LS-Dyna Forum, [13] Snedeker, J., Walz, F., Muser, M., Lanz, C., Assessing femur and pelvis injury risk in car-pedestrian collisions: comparison of full body PMTO impacts, and a human body finite element model, ESV 2005 [14] Yasuki, T., Yamamae, Y., Validation of kinematics and lower extremity injuries estimated by total human model for safety in SUV to pedestrian impact test, Journal of biomechanical science and engineering, Vol5, 2010 [15] Watanabe,R., Katsuhara, T., Miyazaki, H., Kitagawa, Y., Yauki, T., Research of the Relationship of Pedestrian Injury to Collision Speed, Car-type, Impact Location and Pedestrian Sizes using Human FE model (THUMS Version 4), Stapp Car Crash J., 2012 [16] Yasuki, T., Using THUMS for pedestrian safety simulation, AachenerKolloquiumFahrzeug- und Motorentechnik, 2005 [17] Masson, C., Serre, T., Cesari, D., Pedestrian-Vehicle accidentacciedent: analysis of 4 full scale test with PMHS, ESV 2007 [18] Meijer, R., Hassel, E. van, Broos, J. et al, Development of a Multi-Body Human Model that Predicts Active and Passive Human Behaviour, IRC-12-70, IRCOBI conference 2012 [19] Rodarius, C., Mordaka, J., Versmissen, T., Bicycle safety in bicycle to car accidents, TNO report TNO- 033-HM , 2008 [20] Watanabe, R., Katsuhara, T., Miyazaki, H., Kitagawa, Y., Yasuki, T., "Research of the relationship of pedestrian injury to collision speed, car type, impact location and pedestrian sizes using Human FE Model (THUMS Version 4)", Stapp Car Crash J [21] Rodarius, C., Mordaka, J., Versmissen, T., Bicycle safety in bicycle to car accidents, TNO report TNO- 033-HM , 2008 [22] Lubbe, N., "Proposal for an updated BLE test", Toyota Motor Europe internal report 71/75

72 10 Acknowledgment This project is co-funded by the 7th FP (Seventh Framework Programme) of the EC - European Commission DG Research Disclaimer The FP7 project has been made possible by a financial contribution by the European Commission under Framework Programme 7. The ation as provided reflects only the authors view. Every effort has been made to ensure complete and accurate information concerning this document. However, the author(s) and members of the consortium cannot be held legally responsible for any mistake in printing or faulty instructions. The authors and consortium members retrieve the right not to be responsible for the topicality, correctness, completeness or quality of the information provided. Liability claims regarding damage caused by the use of any information provided, including any kind of information that is incomplete or incorrect, will therefore be rejected. The information contained on this website is based on author s experience and on information received from the project partners. 72/75